xlsr-nm-nomimo

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5614
• Wer: 0.3823

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 132
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.9221	4.3478	200	3.0837	1.0
3.0104	8.6957	400	2.8149	1.0
2.6446	13.0435	600	2.2163	1.0
1.605	17.3913	800	0.9394	0.7108
0.6986	21.7391	1000	0.7128	0.5799
0.3653	26.0870	1200	0.5925	0.4724
0.2403	30.4348	1400	0.6623	0.4884
0.1829	34.7826	1600	0.6467	0.4564
0.1415	39.1304	1800	0.6584	0.4462
0.1127	43.4783	2000	0.6751	0.4462
0.0977	47.8261	2200	0.6630	0.4142
0.0794	52.1739	2400	0.5528	0.4230
0.0733	56.5217	2600	0.5641	0.4041
0.0574	60.8696	2800	0.6927	0.4012
0.0518	65.2174	3000	0.6562	0.3983
0.0375	69.5652	3200	0.6104	0.3852
0.0352	73.9130	3400	0.5976	0.3852
0.0297	78.2609	3600	0.6563	0.3852
0.0268	82.6087	3800	0.5655	0.3706
0.0225	86.9565	4000	0.6450	0.3823
0.0213	91.3043	4200	0.6029	0.3837
0.017	95.6522	4400	0.5496	0.3808
0.0166	100.0	4600	0.5614	0.3823

Framework versions

• Transformers 4.47.0.dev0
• Pytorch 2.4.0
• Datasets 3.0.1
• Tokenizers 0.20.0